CN-116369942-B - Electroencephalogram monitoring system and method for neurology

CN116369942BCN 116369942 BCN116369942 BCN 116369942BCN-116369942-B

Abstract

The invention provides an electroencephalogram monitoring system and method for neurology, comprising an electroencephalogram acquisition module, an electroencephalogram processing module, a characteristic extraction module and an electroencephalogram identification module, wherein the electroencephalogram acquisition module is configured to acquire electroencephalogram signals and preprocesses the electroencephalogram signals to acquire first electroencephalogram signals, the electroencephalogram processing module is configured to remove artifacts in the first electroencephalogram signals to acquire second electroencephalogram signals, the characteristic extraction module is configured to acquire characteristic vectors of the second electroencephalogram signals, and the electroencephalogram identification module is configured to output identification results according to the characteristic vectors. According to the invention, by means of independent component analysis and DWT decomposition, the ocular artifacts in the first electroencephalogram signal are precisely removed, and a clean second electroencephalogram signal is obtained. According to the invention, the characteristic vector formed by the sample entropy is identified by adopting the convolutional neural network with a plurality of different activation functions, so that the stability of the system and the accuracy of the identification result are improved.

Inventors

YAN QINGBAO
WANG WEI
XUE SHAN

Assignees

大庆龙南医院

Dates

Publication Date: 20260505
Application Date: 20230428

Claims (6)

1. An electroencephalogram monitoring system for neurology, comprising: the electroencephalogram acquisition module is configured to acquire an electroencephalogram signal, and preprocess the electroencephalogram signal to acquire a first electroencephalogram signal; The electroencephalogram processing module is configured to be used for eliminating artifacts in the first electroencephalogram signals and obtaining second electroencephalogram signals; a feature extraction module configured to obtain a feature vector of the second electroencephalogram signal; The electroencephalogram identification module is configured to output an identification result according to the feature vector; The specific process for acquiring the second electroencephalogram signal comprises the following steps: Performing independent component analysis on the first electroencephalogram signals; determining an independent component containing artifacts according to the detection threshold; Performing DWT decomposition on the independent component containing the artifact; Determining an artifact component according to boundary values of the low-frequency approximate component and the high-frequency detail component of the DWT decomposition; reconstructing the DWT decomposition result after removing the artifact component; reconstructing the reconstructed independent component and the rest independent components without artifacts to obtain a second electroencephalogram signal; performing MEMD (mechanical decomposition) on the second electroencephalogram signal to obtain IMF components after decomposition of each channel; calculating the correlation coefficient of each IMF component and the corresponding channel of the second electroencephalogram signal; calculating sample entropy of IMF components with correlation coefficients larger than a preset threshold value, and forming a feature vector by the sample entropy; The correlation coefficient The calculation method of (1) is as follows: Wherein, the As a component of the IMF, For the corresponding second electroencephalogram signal of the corresponding channel of IMF, Is the length of the signal sequence.
2. The system of claim 1, wherein the specific process of outputting the recognition result according to the feature vector comprises: and inputting the feature vector into a convolutional neural network for recognition, wherein the convolutional neural network respectively adopts different activation functions for recognition and outputs recognition results.
3. The system according to claim 2, wherein: The activation functions are Sigmoid, reLU, and leak ReLU.
4. An electroencephalogram monitoring method for neurology, the method being applied to the electroencephalogram monitoring system for neurology as claimed in any one of claims 1 to 3, comprising the steps of: Step S1, acquiring an electroencephalogram signal, and preprocessing the electroencephalogram signal to acquire a first electroencephalogram signal; s2, performing independent component analysis on the first electroencephalogram signal; step S3, determining independent components containing artifacts according to a detection threshold; Step S4, DWT decomposition is carried out on the independent components containing the artifacts; S5, determining an artifact component according to boundary values of a low-frequency approximate component and a high-frequency detail component of the DWT decomposition; S6, reconstructing a DWT decomposition result after removing the artifact component; S7, reconstructing the reconstructed independent component and the rest independent components without artifacts to obtain a second electroencephalogram signal; performing MEMD (mechanical decomposition) on the second electroencephalogram signal to obtain IMF components after decomposition of each channel; calculating the sample entropy of the IMF component with the correlation coefficient larger than a preset threshold value, and forming a feature vector by the sample entropy; The correlation coefficient The calculation method of (1) is as follows: Wherein, the As a component of the IMF, For the corresponding second electroencephalogram signal of the corresponding channel of IMF, Is the length of the signal sequence.
5. The method according to claim 4, wherein: And inputting the characteristic vector into a convolutional neural network for identification.
6. The method according to claim 5, wherein: the convolutional neural network is respectively identified by adopting different activation functions and outputs an identification result.

Description

Electroencephalogram monitoring system and method for neurology Technical Field The invention relates to the technical field of electroencephalogram monitoring, in particular to an electroencephalogram monitoring system and method for neurology. Background The brain electrical signal records the electrical activity of brain neurons, contains a large amount of physiological information, carries out long-term monitoring and analysis on the brain electrical signal, and has important significance for the research of neurology. In order to improve the accuracy and reliability of signal classification in the brain electrical monitoring process of the neurology, the eye electrical artifact treatment in the brain electrical signal is of great importance. However, the existing brain electrical monitoring system for neurology has the defects that the eye electrical artifacts in the brain electrical signals are not removed accurately, so that the clean brain electrical signals cannot be obtained, the convolutional neural network only adopts a single activation function, so that the identification result is inaccurate, and the system stability is insufficient. Disclosure of Invention Aiming at the problems in the prior art, the invention provides an electroencephalogram monitoring system and method for neurology. In one aspect, the invention provides an electroencephalogram monitoring system for neurology, comprising: the electroencephalogram acquisition module is configured to acquire an electroencephalogram signal, and preprocess the electroencephalogram signal to acquire a first electroencephalogram signal; The electroencephalogram processing module is configured to be used for eliminating artifacts in the first electroencephalogram signals and obtaining second electroencephalogram signals; a feature extraction module configured to obtain a feature vector of the second electroencephalogram signal; And the electroencephalogram identification module is configured for outputting an identification result according to the characteristic vector. Further, the specific process of acquiring the second electroencephalogram signal includes: Performing independent component analysis on the first electroencephalogram signals; determining an independent component containing artifacts according to the detection threshold; Performing DWT decomposition on the independent component containing the artifact; Determining an artifact component according to boundary values of the low-frequency approximate component and the high-frequency detail component of the DWT decomposition; reconstructing the DWT decomposition result after removing the artifact component; And reconstructing the reconstructed independent component and the rest independent components without artifacts to obtain a second electroencephalogram signal. Further, the specific process of obtaining the feature vector of the second electroencephalogram signal includes: performing MEMD (mechanical decomposition) on the second electroencephalogram signal to obtain IMF components after decomposition of each channel; calculating the correlation coefficient of each IMF component and the corresponding channel of the second electroencephalogram signal; and calculating sample entropy of the IMF component with the correlation coefficient larger than a preset threshold value, and forming a feature vector by the sample entropy. Further, the specific process of outputting the recognition result according to the feature vector comprises the following steps: and inputting the feature vector into a convolutional neural network for recognition, wherein the convolutional neural network respectively adopts different activation functions for recognition and outputs recognition results. Further, the activation functions are Sigmoid, reLU, and leak ReLU. In another aspect, the present invention provides an electroencephalogram monitoring method for neurology, the method is applied to the electroencephalogram monitoring system for neurology, and the method includes the following steps: Step S1, acquiring an electroencephalogram signal, and preprocessing the electroencephalogram signal to acquire a first electroencephalogram signal; s2, performing independent component analysis on the first electroencephalogram signal; step S3, determining independent components containing artifacts according to a detection threshold; Step S4, DWT decomposition is carried out on the independent components containing the artifacts; S5, determining an artifact component according to boundary values of a low-frequency approximate component and a high-frequency detail component of the DWT decomposition; S6, reconstructing a DWT decomposition result after removing the artifact component; And S7, reconstructing the reconstructed independent component and the rest independent components without artifacts to obtain a second electroencephalogram signal. Further, MEMD decomposition is carried out on the second electroencephal